The Study Of An ANN Inverse Decoupling Control Of PMSM Motors Based On DSP

Permanent Magnet Synchronous Motors(PMSM) are core parts of many high precision,high efficiency and low power automatic control systems. Improving on the strategies of PMSM control can enhance the performances of those automatic systems directly.This dissertation focuses on some problems in decoupling control strategies,such as that only can realize static decoupling,run short of both the robustness on parameter variation and the ability to resist load disturbance.These strategies depend on Permanent Magnet Synchronous Motor (PMSM) models.Combine neural networks with inverse system method,the neural network inverse system method of Permanent Magnet Synchronous Motor control is proposed.The Permanent Magnet Synchronous Motor,which is a multi-variable,strongly coupling and nonlinear object,is linearised and decoupled into two SISO subsystems.The influence caused by parameter variation and load disturbance decrease evidently.This method provides a new approach for high performance control of Permanent Magnet Synchronous Motor. Main progresses in this dissertation are as follows:First,the decoupling control theory based on inverse system method is studied by using the linear algebraic method of nonlinear systems.The necessary and sufficient conditions for invertibility of nonlinear system are developed,and the construction algorithms for pseudo-linear subsystems are given respectively.Secondly,the invertibility and decoupling property of Permanent Magnet Synchronous Motor are analyzed systematically and thoroughly.The structures of the state feedback combined with the input integral inverse system,which achieve linearization and decoupling of Permanent Magnet Synchronous Motor, are put forward.The influence of motor parameter variation and load disturbance to the decoupling control performance is discussed,which shows that the analytical inverse method is not able to achieve the high performance for Permanent Magnet Synchronous Motor control.In order to eliminate the influence resulted from parameter variation and load disturbance,a neural network inverse system method is proposed to realize the decoupling control of Permanent Magnet Synchronous Motor.The structure of neural network inverse system,the identification approach and realization steps to obtain the neural network inversion are given.After the neural network inversion is connected before the Permanent Magnet Synchronous Motor in series,the Permanent Magnet Synchronous Motor is decoupled into two SISO pseudo-linear integral subsystems.Then,close-loop linear controllers are designed.The control system has good robustness to parameter variation and strong adaptability to load disturbance.Finally,the proposed method is validated through a controlling experimental platform which is based on digital signal processor(DSP) and intelligent power module(IPM).The control performance is satisfying,which shows that the neural network inverse system method is an effective and applicative method for the control of Permanent Magnet Synchronous Motor.